Staffing Problems

[Cirep10]

Hello Everyone,

From working in the call center industry for a couple of years I have noticed that staffing will be a continuous battle. Well, I am sending this new thread today hoping someone could shed a little light to my dilemma. I

I have utilized the Erlang calculators so being the great toll that it is. It has given me an idea how many agents I would need in an hour to handle a certain call volume. My problem is knowing how many agents I would need for a entire work day working off of 7.2 man hours (includes breaks, lunch, DND, vacations sick time).

I work for an outsourcing company that administers matching gift, giving campaigns and also volunteer programs for over 250 companies and there are currently 250 separate phone lines for each client we administer the program for. I currently have a 12-man team with 2 part-time employees and 10 full-time employees. We average about 1000 calls per day in our busy season which is Sept - Jan. If anyone can help me please help...

[JoeB]

When call load is uniform during the day, you can use Erlang to calculate staffing per hour, which will be the same throughout the day. If call load fluctuates, such as a different call volume during early morning, around noontime, etc., then you need to calculate staffing for each of these periods separately, and roster the shifts accordingly.

As for the time you lose for breaks, vacation, etc., which we refer to as shrinkage, you need to increase the headcount proportionally. Some Erlang calculators (e.g. http://www.diagnosticstrategies.com/EasyErlang.htm) calculate shrinkage adjustments for you.

For manual calculation, the simplest approach is to increase headcount by the percentage of time lost to shrinkage, e.g. multiply headcount by 1.2 for 20% shrinkage. Bearing in mind that the additional 20% will also be subject to the same shrinkage, a safer approach is to multiply headcount by the shrinkage squared, e.g. 1.44.

See more about shrinkage calculations in http://www.diagnosticstrategies.com/....htm#Shrinkage

[forc3d]

Originally posted by JoeB
When call load is uniform during the day, you can use Erlang to calculate staffing per hour, which will be the same throughout the day. If call load fluctuates, such as a different call volume during early morning, around noontime, etc., then you need to calculate staffing for each of these periods separately, and roster the shifts accordingly.

As for the time you lose for breaks, vacation, etc., which we refer to as shrinkage, you need to increase the headcount proportionally. Some Erlang calculators (e.g. http://www.diagnosticstrategies.com/EasyErlang.htm) calculate shrinkage adjustments for you.

For manual calculation, the simplest approach is to increase headcount by the percentage of time lost to shrinkage, e.g. multiply headcount by 1.2 for 20% shrinkage. Bearing in mind that the additional 20% will also be subject to the same shrinkage, a safer approach is to multiply headcount by the shrinkage squared, e.g. 1.44.

See more about shrinkage calculations in http://www.diagnosticstrategies.com/....htm#Shrinkage

Say, I'd like to compute my shrinkage manually...How far should i backtrack to arrive at a more realistic value, considering that I'm deploying at least 25 new agents per month? Should I be including attrition in my shrinkage computation, or is this treated separately? Should I also be considering shrinkage per interval? Any help is much appreciated.

Best.

[JoeB]

I think that attrition should not be factored into shrinkage calculations. Shrinkage is designed to indicate unplanned and “localized” reduction in manpower availability.

Unless shrinkage is varies dramatically from one interval to another, calculating it for each is unnecessary. Just add the shrinkage to each shift or staffing interval.

A general note about the forgoing and previous discussion. I believe that while the approach to call center staffing should be based on sound principles and mathematical tools exists to model the operation, plan, and forecast performance, over reliance on these calculations can be risky. Use these calculations as a starting point and monitor performance and shrinkage closely, adapting the models to your specific environment.

Joe

[n_meghanath]

Staff forecast can only be accurate once you have an accurate call volume forecast. Assuming you have done a good homework on the trend analysis of the call flow and you have derive the near to accuracy probable call volume interval wise you might even use excel to find the required staffing. Another factor for staffing would be the trend of your AHT (Average Handle time). or you may take the target AHT for calculation depending on the feasibility of maintaing under the target. Now using a simple calculation you may find yourself doing wonderfully well with the staff forecast.

Taking a call volume of 50 in an interval of 30 mins with AHT target of 300 sec and a target of 70% telephone occupancy you may calculate:

time required to answer 50 calls=50*300=15000
staff time required@ 70% T occupancy=15000*100/70=21428
Required staff to ans 50 calls in an interval of 30 mins=21428/30*60=12 (rounding of 11.9)

So you should have 12 agents during this 30 mins interval to handle all the 50 calls and plus the shrinkage

say if you are taking a shrinkage of 12% depending on what you include in the shrinkage. then your staffing would be 12+1.44=13.44 you may take either 13 or 14 to be on the safer side.

It may not be very reliable but it does work where there is no much fluctuation on the aht and call volume.

[JoeB]

Unfortunately this ”basic” approach is incorrect. Mathematically speaking, you could argue that 30% shrinkage is not adjusted for correctly when multiplying the available time by 70% but rather by dividing it, but these differences are not likely to have pragmatic differences.

But more importantly, the assumption that in order to figure out headcount you simply divide the call load (call volume * AHT) by the “call clearing rate” is incorrect. The reason for that is that calls do not arrive in an orderly fashion, such that as soon as the agent finished one call the next caller arrives. The reality is that sometimes several calls arrive at once, or there are very few calls, and other times there are no calls. Said another way, AHT of 300 seconds does not mean a call every 300 seconds. In fact, most of the time calls will arrive much faster than that, and an interesting point is that with AHT of 300 seconds, the higher probability is to get calls slightly faster than once every 300 seconds.

Just as an illustration, the number of agents needed to support the call load and shrinkage, and service level of 80/20 in the example is actually 16

Please read the article http://www.diagnosticstrategies.com/...c_modeling.htm for a detailed description of the background and the correct way to calculate staffing.